One pattern I find useful for working with LLMs is a nice long ramble session. Sometimes the LLM needs more bits to understand what you're trying to achieve, but you're too lazy to type them. In these cases I like to lean back, switch to /voice and just ramble for like 10 minutes, total mess, anything goes, full stream of consciousness. Sometimes I declare it up top, something like "switching to speech recognition sorry for any typos...". Sometimes I turn it into a small interview of a few turns. But I find that the LLMs are somehow very good at reconstructing long incoherent rambles and often their echo of your own tangle of thoughts comes out quite a bit cleaner than what you started with. The result is that you improve the mind meld and have to correct things less from that point on.
我发现一个对付 LLMs 很有用的模式，是来一场又长又随意的 ramble（漫谈）session。有时候，LLM 需要更多信息碎片来理解你到底想达成什么，但你又懒得把这些都敲出来。这种情况下，我喜欢往后一靠，切到 /voice，然后就开始一通讲上 10 分钟左右，完全乱来，想到哪说到哪，彻底的 stream of consciousness（意识流）。有时我会先在开头声明一下，比如说“切到 speech recognition 了，抱歉如果有任何 typos……”。有时我会把它变成一个只有几轮的小采访。但我的体会是，LLMs 不知怎么就特别擅长重建这种又长又不连贯的漫谈，而且它们对你那一团乱麻式思路的“回声”，往往会比你最初说出来的东西干净清楚不少。结果就是，你和模型之间的 mind meld（思维对齐）会做得更好，之后需要纠正它的地方也会少很多。